change-review
Review a PR, branch or local diff before it merges.
How do I install this agent skill?
npx skills add https://github.com/repowise-dev/repowise --skill change-reviewIs this agent skill safe to install?
- Gen Agent Trust Hubpass
The skill provides instructions for an AI agent to perform automated code and security risk reviews using Repowise tools. It is generally safe but carries a low risk of indirect prompt injection as it is designed to process untrusted diff data from pull requests and branches.
- Socketpass
No alerts
- Snykwarn
Risk: MEDIUM · 1 issue
What does this agent skill do?
Change Review with Repowise
When a diff is on the table, Repowise turns "what files changed" into "what does this change put at risk" — fusing git history (churn, ownership, co-change) with graph topology (dependents, impact surface), test gaps, security signals, and the architectural decisions that govern the touched code.
Two complementary risk signals, use both:
get_change_risk(revspec=…)(MCP) scores the whole change as one unit (a commit or abase..headrange). No LLM, no network. Prefer this in-MCP tool; it takes a revspec and diffs server-side, so you never shell out. Lead withdirectiveandhealth_delta— what the change actually made worse. Findings the change wrote sort above pre-existing ones it only touched. Thenrisk_percentile(this change ranked against sampled recent commits by diff size and spread, not a bug predictor), stated in one line bydiff_shape;include=["scales"]adds itsreview_prioritytercile andclassificationlabel. The raw 0-10scoreranks 0.99 against lines added, so it sits behindinclude=["diagnostics"]rather than on the wire. Omitrevspecto score uncommitted work. This is the pre-merge gate: "how risky is this change overall?" Therepowise risk <revspec>CLI is the identical scorer for when you are already in a terminal.get_risk(changed_files=…)(MCP) works per file and returns thedirectiveblock, the specific things to check inside the diff.
Score the whole change first
get_change_risk(revspec="main..HEAD") # HEAD, a commit SHA, or base..head
Read directive first: status, a headline naming what got worse, the
reasons behind it and next_actions to run. health_delta carries the
findings the change introduced or resolved. Then risk_percentile and the top
drivers: a high score from large diffusion (many dirs/subsystems) or low author
familiarity tells you where to look hardest. extensions=[".py", ".ts"] counts
only certain file types; exclude_patterns=["tests/"] omits paths. A warning
field, or a directive.status of unknown, means the revspec or filters
matched no files, so an all-zero score there is not a clean bill of health. The
equivalent from a terminal is repowise risk <revspec> (add --ext .py,.ts or
--format json).
Then drill into the per-file directive
Call get_risk in PR mode by passing the changed files:
get_risk(targets=<changed files>, changed_files=<same changed files>)
The response carries its own directive block, per file rather than per change
(distinct from get_change_risk's) — read it first, it's a few short lists:
may_break— files/symbols that import their way to what changed but are not in the diff. Reachability, not a diff: it says these depend on the changed files, not that the thing they use changed. Check each one.missing_cochanges— files that historically change together with the changed files but were left untouched. Often a forgotten update.missing_tests— changed code with a test gap. Flag for new/updated tests.tests_to_run: the tests the change needs, from the same selectionrepowise impacted-testsmakes, withtests_to_run_whygiving each one's reason (the changed file and its evidence, or "runs with every subset"). Checktests_run_allfirst: when it is true no smaller set can be vouched for,tests_run_all_reasonssays why, and the list is only what to run first; recommend the whole suite. When it is false the list is the set to run.tests_to_run_basisismeasuredwhen stored per-test coverage decided a changed file andinferredwhen only the dependency graph or a rule did; say which.tests_statusappears only when no selection could be made (timeout,config_invalid,unknown, ...), and then every test must run.
get_change_risk carries the same answer as impacted_tests: status
(selected or run_all, or why nothing was selected), run_all, reasons,
tests_to_run and why. include=["blast"] on get_risk adds
pr_blast_radius, the fuller dossier behind the directive's lists.
For the line-precise version from a terminal, repowise impacted-tests <revspec>
maps each changed line to the tests whose recorded coverage touches it, then
prints the ids (--format list | xargs pytest runs exactly them). It is honest
about gaps: a changed file with no coverage rows is a labelled filename guess,
and a brand-new file is "unknown, run the full suite", never "no tests needed".
Then go deeper where it matters
- Why does this code exist? For any non-trivial changed file, call
get_why(query="<file>")— don't let a change silently contradict a recorded architectural decision. Surfaceconflicts_with/supersedeshits. - Did the change make health worse?
get_health(targets=<changed files>, include=["biomarkers"])— call out new complexity, deep nesting, or duplication the diff introduced. - Who should review?
get_riskownership + co-change signals suggest the people with the most context on the touched code.
Getting the diff
- A GitHub PR:
gh pr diff <number>(orgh pr view <number> --json files). - A branch:
git diff --name-only main...HEAD. - Working tree:
git status --porcelain. - CLI shortcut for a range:
repowise risk main..HEADscores a branch/PR range for defect risk directly.
Write the review around evidence
Lead with a risk level and the directive findings, each tied to a concrete
file. Distinguish "may break" (a dependent outside the diff) from "worth a
look" (a co-change or health regression). Don't pad with findings the tools
didn't support.
Error handling
If get_risk errors or returns nothing, the MCP server may be down or the repo
unindexed — say so and review from the raw diff, noting that Repowise context was
unavailable. Suggest /repowise:init if the repo isn't indexed.
How can the creator link this skill?
Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.
<a href="https://skillzs.dev/skills/repowise-dev/repowise/change-review">View change-review on skillZs</a>